Predictors of New Need for Postoperative Hormone Supplementation After Nonfunctional Pituitary Neuroendocrine Tumor Resection: A Retrospective Review of 701 Cases

Introduction: Postoperative hypopituitarism is one of the most common complications after surgical removal of pituitary neuroendocrine tumors (PitNETs). A better understanding of which patients are at increased risk for this complication would inform discussion of surgical risks and postoperative planning.

Methods: A retrospective review was performed on patients undergoing surgical removal of nonfunctional PitNETs from 1994 to 2019 at a single high-volume academic pituitary center. Hypopituitarism was declared when patients demonstrated a new postoperative need for exogenous hormone supplementation therapy more than 3 months after surgery. Hormonal supplementation was based on published guidelines by a single expert neuroendocrinologist. Covariates examined included patient sex, age at surgery, tumor size, cavernous sinus invasion, number of prior pituitary surgeries, preoperative hormone and sodium levels, extent of resection, intraoperative or postoperative cerebrospinal fluid (CSF) leak, and postoperative diabetes insipidus. Variables’ effects on postoperative hypopituitarism were examined by Pearson’s chi-squared tests, Wilcoxon’s rank sum tests, and Fisher’s exact tests, where appropriate. A random forest machine learning model was built to assess for variable importance in determining hypopituitarism status. Preoperative variables found to be significant hypopituitarism predictors were subsequently assessed via receiver operating characteristic (ROC) analysis to determine numeric thresholds that optimally predicted preoperative likelihood of postoperative hypopituitarism.

Results: A total of 701 patients were identified with nonfunctional PitNETs who underwent transsphenoidal resection, with 35.0% (n = 245) experiencing new postoperative hypopituitarism after a median follow-up of 21 months. Statistical testing revealed preoperative FSH to significantly predict postoperative hypopituitarism in women (odds ratio 0.96 per serum mIU/mL, p = 0.008). A random forest machine learning model with 10,000 iterations found important preoperative predictors to include maximum tumor dimension as well as preoperative serum FSH, prolactin, and cortisol ([Fig. 1]). ROC analysis revealed patients with tumors >2.05 cm were more likely to experience postoperative hypopituitarism (76.5% sensitivity, p = 0.001 for men, [Fig. 2]; 64.5% sensitivity, p = 0.029 for women). For females, 91.9% of patients with intact postoperative hormone function had preoperative serum FSH greater than 27.95 mIU/mL (AUC = 0.692, p < 0.001; [Fig. 3]).

Conclusions: 77% of men and 65% of women with postoperative hypopituitarism had preoperative tumors larger than 2 cm. Preoperative low prolactin and cortisol levels may affect risk of hypopituitarism in all patients. Women with preoperative serum FSH > 27.95 mIU/mL are unlikely to experience hypopituitarism after surgery.

Fig. 1 Results of a random forest machine learning model to predict postoperative hypopituitarism. Variables are scored by mean decrease in model accuracy when the variable is removed from the model. A higher score indicates increased importance.

Fig. 2 ROC analysis of preoperative tumor size predicting postoperative hypopituitarism in men.

Fig. 3 ROC analysis of preoperative FSH predicting normal postoperative endocrine function in women.

Publikationsverlauf

Artikel online veröffentlicht:
05. Februar 2024